RAG & Retrieval · Ground answers in evidence
RAG vs fine-tuning: how do you choose?
Short answer
Use RAG for knowledge that changes, needs citations or varies by user. Use fine-tuning for stable behaviour, style or format. They solve different problems and often combine.
RAG injects knowledge at request time. Fine-tuning changes the weights so a behaviour becomes more consistent. A model fine-tuned on last quarter's policies will still be wrong this quarter, and a RAG system cannot teach a model a new tone reliably. Match the intervention to the failure.
- Facts that change often, or must be cited: RAG.
- Consistent tone, output format or narrow task behaviour: prompt first, fine-tune if needed.
- Access control by user or tenant: RAG, because retrieval can filter documents.
- Long prompts used only to teach style: fine-tuning may reduce cost at scale.
Follow-ups you should expect
Interviewers use these to check whether the first answer was understanding or memorisation.
- 1How would you evaluate whether fine-tuning helped?
- 2Why is RAG better for rapidly changing policy documents?
- 3Can you combine both? What would each part do?
Say it in the interview
I use RAG when the problem is knowledge: it changes, it needs citations, or access differs by user. I use fine-tuning when the problem is behaviour: tone, format or a narrow task the model keeps getting subtly wrong. Fine-tuning cannot keep facts current, and RAG cannot reliably reshape style. In practice I might fine-tune for format and use RAG for the content, then evaluate the combination.
Use this as a model, not a script. Saying it in your own words is what makes it survive follow-up pressure.
Related questions
Reading is step one
Now say it out loud.
Understanding an answer and delivering it under pressure are different skills. Rehearse this topic in a live RAG & Retrieval mock interview and get scored feedback on depth, clarity and confidence.